Hyperbolic cleaning damps psi once per spatial direction #1993

In s_compute_rhs (src/simulation/m_rhs.fpp, the if (hyper_cleaning) block around line 762), the damping term

rhs_vf(eqn_idx%psi)%sf(j, k, l) = rhs_vf(eqn_idx%psi)%sf(j, k, l) - q_prim_vf(eqn_idx%psi)%sf(j, k, l)/hyper_cleaning_tau

sits inside the dimensional-splitting loop do id = 1, num_dims (line ~683). It is therefore added num_dims times per RHS evaluation, so the effective damping time is hyper_cleaning_tau/num_dims rather than hyper_cleaning_tau (halved in 2D, a third in 3D; hyper cleaning is prohibited in 1D).

Expected: the source term -psi/tau is applied once per RHS evaluation. Moving the block out of the id loop (after END: Dimensional Splitting Loop) would do that. This changes results for every hyper_cleaning = T case, so the hyper-cleaning goldens (B4DC99F9, EC8F87D9, F057F8E6) and possibly the recommended hyper_cleaning_tau values in examples/*hyper_cleaning* would need revisiting.

Related: #1987 fixes which psi value the damping reads (stale q_prim_vf vs. current-stage q_prim_qp), but keeps the term inside the loop.